I never imagined I would one day share a speaking session — and a speaker room chat — with the 𝗹𝗲𝗴𝗲𝗻𝗱𝗮𝗿𝘆 𝗔𝘂𝘀𝘁𝗿𝗮𝗹𝗶𝗮𝗻 𝗰𝗿𝗶𝗰𝗸𝗲𝘁 𝗰𝗮𝗽𝘁𝗮𝗶𝗻 𝗦𝘁𝗲𝘃𝗲 𝗪𝗮𝘂𝗴𝗵. Thanks to the Zoho event in Sydney, that happened. Speaking right after Hugh Watson Australia’s Ambassador for Cyber Affairs and Critical Technology also added a valuable global perspective to the discussion.

In my fireside chat, I reflected on our journey helping develop Australia’s AI safety standard: 𝗽𝗿𝗮𝗰𝘁𝗶𝗰𝗮𝗹 𝗴𝘂𝗶𝗱𝗮𝗻𝗰𝗲 𝗳𝗼𝗿 𝗦𝗠𝗘𝘀 𝗮𝗻𝗱 𝗔𝗜 𝗱𝗲𝗽𝗹𝗼𝘆𝗲𝗿𝘀 𝘁𝗵𝗮𝘁 𝗻𝗼𝘁 𝗼𝗻𝗹𝘆 𝗳𝗶𝘁𝘀 𝘁𝗵𝗲 𝗔𝘂𝘀𝘁𝗿𝗮𝗹𝗶𝗮𝗻 𝗰𝗼𝗻𝘁𝗲𝘅𝘁 𝗯𝘂𝘁 𝗮𝗹𝘀𝗼 𝗮𝗹𝗶𝗴𝗻𝘀 𝗰𝗹𝗼𝘀𝗲𝗹𝘆 𝘄𝗶𝘁𝗵 𝗶𝗻𝘁𝗲𝗿𝗻𝗮𝘁𝗶𝗼𝗻𝗮𝗹 𝗳𝗿𝗮𝗺𝗲𝘄𝗼𝗿𝗸𝘀.

On adoption barriers, we discussed an inconvenient truth: AI adoption is everywhere at the individual level — from personal life to employees secretly experimenting or building tools for colleagues in unsanctioned ways. Yet it struggles to bubble up to the business level.

I do believe AI assurance and testing are critical for broad adoption. But the real challenge is this: how do we harness the innovation power of individuals without being paralysed by fear?

One way forward is sensible evaluation before deployment, backed by strong post-deployment monitoring and assurance. At CSIRO’s Data61, we are developing such a balanced approach that spreads risk management across DevOps pipelines rather than relying solely on large upfront risk assessments. That means: resisting the urge to make systems “perfect” before release, because this wave of GenAI won’t wait.

Finally, I was asked how I personally use AI. The timing couldn’t be better — this Friday I’ll deliver the second instalment of my talk at another APS event, building on my recent (slightly controversial) reflections in 𝗥𝗲𝗮𝗱𝗶𝗻𝗴 𝗶𝗻 𝘁𝗵𝗲 𝗔𝗴𝗲 𝗼𝗳 𝗔𝗜. This time, I’ll go even further, tackling a more challenging topic — 𝗪𝗿𝗶𝘁𝗶𝗻𝗴 𝗶𝗻 𝘁𝗵𝗲 𝗔𝗴𝗲 𝗼𝗳 𝗔𝗜 — where I’ll make another controversial point. Stay tuned.